A hierarchical synthetic aperture radar image registration method for change detection

The synthetic aperture radar (SAR) images obtained at different times often have both the global rigid deformation and the local elastic deformation. As a result, SAR image registration only based on the global rigid transformation model will induce an error, and the error will produce largely spurious results of change detection. In order to solve this problem, a hierarchical registration method is proposed in this chapter. It is a rough-to-fine registration procedure. In the rough registration stage, speeded-up robust reature (SURF) algorithm is first employed to extract the matching feature points from the testing image and the reference image. After that, a rigid registration method, based on the global affine transformation model, is implemented to correct the global deformation. In the fine registration stage, a template-match-based method is first adopted to extract a uniform spacing control point from the globally registered images. And then, a local elastic registration, based on thin-plate spline (TPS) transformation model, is applied to remove the local deformation. Experimental results show that the proposed approach can achieve higher registration accuracy, which lead to a better change-detection performance. © Springer International Publishing Switzerland 2015.